The Ups and Downs of EEG. Electroencephalography in Functional Asymmetry Research

Table of Contents

1 Introduction

2 Methods of brain asymmetry research
2.1 Functional brain asymmetry
2.2 Electroencephalography in functional asymmetry research

3 Electroencephalography applied to topics of functional brain asymmetry research
3.1 Emotions
3.2 Language processing and the N1 component
3.3 Face processing and the N170 component

4 Limitations of EEG
4.1 Spatuial resolution
4.2 Causation
4.3 Choosing an appropriate research method (ChARM) model

5 Concluding remarks

References

Abstract

A variety of methods is available for today’s functional brain asymmetry research. After giving a short overview of methods especially used in former times (e.g., lesion studies and microstimulation in patients), this review introduces the method of electroencephalography (EEG) and discusses its advantages. For instance, those include EEG’s high temporal resolution and its relatively low costs of purchase and maintenance. Furthermore, the review elaborates EEG’s contributions to functional asymmetry research topics such as the processing of emotions, language, and faces. Additionally, by presenting research conducted with alternative methods, we can show that they yield comparable or complementary findings. It becomes clear that while EEG has its merits, a combination of methods is integral in order to thoroughly investigate brain asymmetry. That is a reason why in the end, a model is suggested in order to provide guidance in choosing an appropriate research method (ChARM) for a specific research question.

1 Introduction

Thanks to the neurologist Broca (1865) who concluded that speech production is lateralized, brain asymmetry is one of the most fundamental premises of today’s neuroscience. Due to the absence of appropriate technological methods for measuring or manipulating cerebral activity, during Broca’s time findings in brain asymmetry research were mainly built upon the structural analysis of post-mortem brains (e.g., Wernicke 1974).

However, lesion research suffers from critical shortcomings. For example, when a functional deficit follows a lesion, one cannot determine if the damaged area is responsible for the function or if the area is part of a larger network (Farah 1994; Rorden and Karnath 2004). Additionally, since compensatory plastic reorganizations of the brain following lesions may take place (Desmurget, Bonnetblanc, and Duffau 2007), it is difficult to control the effect of these reorganizations on the observed post-lesion functions.

As an example of an alternative method, Penfield and Rasmussen (1950) inserted electrodes in patients‘ brains to stimulate their motor and somatosensory cortices, in order to observe the link between microstimulation and the subject’s response. However, because of its invasiveness, Penfield and Rasmussen (1950) did not apply microstimulation in healthy humans but in patients suffering from epilepsy, prior to their operation.

EEG is a method suited to tackle these shortcomings. Berger (1929) first used EEG to record electrical waves from an intact human cortex, allowing for measurement of normal cortical activity. Since then, a huge amount of functional brain asymmetry research has been based on EEG. The goal of this paper is to give an overview of the current EEG asymmetry research, describing EEG’s advantages and limitations compared to other available methods. Finally, it will recommend methods that are suitable for particular future research questions regarding brain asymmetry.

2 Methods of brain asymmetry research

2.1 Functional brain asymmetry

Functional and structural brain and behavioral asymmetries have been detected in humans and animals (Toga and Thompson 2003). For instance, Broca (1865) and Wernicke (1974) analyzed the language abilities and post-mortem brains of their patients. However, these procedures do not cover research on functional brain asymmetries since the two researchers investigated non-functioning brains. In contrast, Hugdahl and Westerhausen (2010) describe functional brain asymmetries as hemispheric differences in processing different aspects of stimuli or tasks. By understanding how EEG operates, it becomes clear why it is suitable for the analysis of functional asymmetries.

2.2 Electroencephalography in functional asymmetry research

2.2.1 Procedure and measures

Researchers measure EEG data by typically attaching between 32 and 128 recording electrodes to the participants’ scalps using an EEG cap. Additionally, a ground and a reference electrode are needed (Ocklenburg and Güntürkün 2017). Jackson and Bolger (2014) furthermore suggest using electrode gel to enhance the electrical signals reaching the recording electrodes. These electrodes measure synchronized electrical activities of populations of cortical neurons (da Silva 2010; Jackson and Bolger 2014). Then, those electrical activities are summed up in an overall EEG signal (da Silva 2010; Jackson and Bolger 2014; Nisar and Yeap 2015). Since the recorded EEG signal does not exceed the micro-volt range, one can use an EEG amplifier to enlarge it (Jackson and Bolger 2014).

Finally, a recording computer measures the raw EEG signal (Ocklenburg and Güntürkün 2017). Subsequent analyses mainly identify oscillations and event-related potentials (ERP) (Ocklenburg and Güntürkün 2017). Oscillations are a product of millions of neurons firing in a synchronized manner and establishing different frequency bands, i.e. rhythmic patterns of action potentials. The most common frequency bands are the Delta-, Theta-, Alpha-, Beta-, and Gamma-band (Nisar and Yeap 2015). Research on hemispheric asymmetries has often relied on alpha oscillations. As those oscillations mainly occur in a relaxed state or with eyes closed, locally observed high alpha power is interpreted as low brain activity in this location (Ocklenburg and Güntürkün 2017).

Tomarken et al. (1992) suggested inferring brain activation from measured inactivation using the alpha power due to relative temporal stability and internal consistency of the Alpha-band within participants, allowing for interindividual comparisons. However, Hagemann (2004) questioned this approach due to methodological difficulties in measuring the Alpha-band. ERPs, on the other hand, are the sum of voltage fluctuations in the raw EEG signal that are time-locked to an event (Kappenman and Luck 2012). EEG is therefore the method of choice when attempting to measure timing-sensitive processes such as oscillations and ERPs, as it has a high temporal resolution (Brancucci 2010; Ocklenburg and Güntürkün 2017).

2.2.2 EEG’s advantages in functional asymmetry research

Especially in the 80s and 90s, EEG was used in many areas of functional asymmetry research, for example emotion processing (Ahern and Schwartz 1985). The main reason for this was the lack of alternative methods. Later alternatives include functional magnetic resonance imaging (fMRI), which was first used in an experimental setting in 1991 (Belliveau et al. 1991), and magnetoencephalography (MEG), first used around 1985 (Noohi and Amirsalari 2016). Many researchers use these methods in today’s asymmetry research (e.g., Baas, Aleman, and Kahn 2004; Xu, Liu, and Kanwisher 2005). Considering this, is there still a need for EEG today?

The EEG’s ability to measure oscillations and ERPs allows for the investigation of neural and mental processes in the millisecond range (Brancucci 2010; Ocklenburg and Güntürkün 2017), which is much higher than the fMRI’s temporal resolution (Luck 2014). While MEG’s temporal and spatial resolution is generally comparable to EEG’s, they differ in the neural information they record (for further information, see Malmivuo 2012). Additionally, EEG is comparably inexpensive (Brancucci 2010; Davidson 1988).

Concerning the importance of EEG’s temporal resolution, the visual N1 is of interest – a negative component of an ERP occurring around 100 to 200 milliseconds after stimulus onset. It has been associated with the processing of all visual stimuli while varying in amplitude depending on stimulus category (Rossion and Jacques 2012). As we will see later, the N1 and similar components measurable by EEG show hemispheric differences in timing or amplitude, indicating hemispheric differences in information processing (e.g., Caharel et al. 2009; Spironelli and Angrilli 2009).

3 Electroencephalography applied to topics of functional brain asymmetry research

3.1 Emotions

According to Gross and Thompson (2007), emotions provide a readiness potential to show a particular reaction to a stimulus. Additionally, the authors suggest three main emotional components: a physiological and a behavioral response to a given or upcoming situation as well as the individual’s subjective experience. Emotions mainly occur when a situation is relevant for the individual’s own goals, they facilitate decision-making and highlight important events with the label “to be remembered” (Gross and Thompson 2007; Roozendaal, McEwen, and Chattarji 2009). Ekman and Friesen (1971) found cross-cultural recognition of six human emotions (happiness, fear, anger, sadness, disgust, and surprise). Therefore, these were later suggested to be the six basic emotions (e.g., Sauter et al. 2010). Although this notion led to a vigorous debate throughout the last century (e.g., Ekman 1994; Russell 1994), modern research on hemispheric asymmetries in emotion processing still refers to these findings (Ocklenburg and Güntürkün 2017).

Two models have been repeatedly suggested when considering hemispheric asymmetries in emotion processing: the right-hemisphere (RH) model and the valence model, which are also the oldest ones (Alves, Fukusima, and Aznar-Casanova 2008; Demaree et al. 2005; Ocklenburg and Güntürkün 2017). Whereas the former states an almost exclusive right-hemispheric dominance in emotion processing, the latter claims a valence-specific hemispheric specialization, with the left hemisphere being involved in positive and the right hemisphere in negative emotion processing. There is evidence for both models. The finding that patients with right-hemispheric lesions display a decrease in emotional expression and often become emotionally indifferent or manic supports the RH model (Alves, Fukusima, and Aznar-Casanova 2008; Demaree et al. 2005). EEG research in healthy subjects confirmed that at least two aspects of emotion processing, perception and recognition of facial expressions and affective prosody, are mostly confined to the right hemisphere (Kestenbaum and Nelson 1992; Laurian et al. 1991, Pihan, Altenmüller, and Ackermann 1997).

Despite this, Goldstein’s (1939) observation that patients with left-hemispheric frontal lesions often display severe symptoms of depression, whereas patients with right-hemispheric lesions often show an increased positive emotional condition, challenged the RH model (as cited in Ocklenburg and Güntürkün 2017). Later on, Robinson et al. (1984) confirmed this concept, but replication problems made Singh, Herrmann, and Black (1998) call it into question. Nevertheless, this finding was undoubtedly one source of inspiration for the valence model, which was further confirmed in behavioral (Ahern and Schwartz 1979; Reuter-Lorenz and Davidson 1981) and EEG studies (Ahern and Schwartz 1985; Davidson and Fox 1982). Bryden (1982) proposed a variant of the valence model stating that valence-specific activations occur in case of emotional expressions, but recognition and reception of emotion is confined to the right hemisphere (which works as an integration of the RH and valence model). Davidson et al. (1979) provided some sort of anticipatory evidence in their EEG study. They found a valence-specific, frontal dissociation of hemispheric activation, but only a rightsided activation in the parietal lobe, irrespective of valence.

Throughout the years, Davidson (1995) expanded the valence model by claiming not only the hemispheric lateralization of valence, but also of motivational aspects of emotions (as cited in Ocklenburg and Güntürkün 2017). This was the dawn of the approach-withdrawal model that states that the processing of negative or withdrawal-related and positive or approach-related emotions is confined to the right and left hemisphere, respectively (as cited in Demaree et al. 2005). Sutton and Davidson (1997; 2000) furthermore demonstrated that an individual’s dispositional anterior resting brain activity was associated with the dispositional motivational direction (i.e. right-hemispheric activation and withdrawal versus left-hemispheric activation and approach, respectively) and predicted the evaluation of emotional stimuli. A criticism of the approach-withdrawal model relates to the emotion of anger. Anger triggers approach behavior despite being a negative emotion and elicits a relative increase in left- compared to right-frontal activation (Harmon-Jones and Allen 1998; Harmon-Jones and Sigelman 2001). The authors concluded that prefrontal asymmetries in emotional processing rather relate to the motivational direction of an emotion than to its valence.

To sum up, all these models have their own strengths and limitations and for each of them, EEG findings added partial support.

3.2 Language processing and the N1 component

Another topic of functional asymmetry which is investigated through EEG is language processing. A fast, left-sided N1 response originating from the occipito-temporal cortex is repeatedly reported in a number of studies comparing word to non-word processing (Maurer, Brandeis, and McCandliss 2005; Maurer et al. 2005; Spironelli and Angrilli 2009), in which N1 was usually identified either as N150 (e.g., Cohen et al. 2000) or N170 (e.g., Bentin et al. 1999). McCandliss, Cohen, and Dehaene (2003) suggest that this left-sided N1 response is an indicator of automatic reading.

The specialization of the left hemisphere for language-related stimuli becomes obvious when looking at a study conducted by Gros et al. (2002): when the letter “o” is presented either with other letters or with figures, the EEG responded to the “o” with the same left N170 response other letters induced in the “letter” condition and with the same right N170 response the figures induced in the “figure” condition. Thus, the context the letter “o” was shown in determined the type of N170 response. Another study by Brem et al. (2005) found that a nonsensical letter string at first did not result in the same left N1 response as in the word condition. But after repetition of the nonsensical letter string, a left N1 response comparable to the word condition occurred. This indicates that perceptual experience and stimulus expertise influence the N1 response. Certain word characteristics also influence the exact duration and source location of the N1 response, such as length and frequency (Hauk and Pulvermüller 2004), or semantic category such as animals and verbs (Dehaene 1995) or color and form (Moscoso del Prado Martín, Hauk, and Pulvermüller 171 2006).

3.3 Face processing and the N170 component

Rossion et al. (2003) observed a specific N170 activation whose origin depended on the stimulus: while word processing led to an N170 in the left occipito-temporal cortex, face processing elicited a right-sided N170 in the contralateral region. Other studies also showed a connection between face perception and an exceptionally strong, right-sided N170 activation compared to non-face stimuli (e.g., Eimer 2000; Goffaux et al. 2003; Itier and Taylor 2004). Additionally, the N170 component is stronger when seeing faces with emotional expressions such as happiness or anger (Hinojosa, Mercado, and Carretié 2015). Generally speaking, research indicates that the right-sided N170 peak reflects the mental process of detecting a face, i.e. the categorization of the stimulus as a face (Bentin et al. 1996; Jeffreys 1996; Rousselet, Macé, and Fabre-Thorpe 2004). Other studies imply that the N170 also reflects individual face discrimination by showing that face adaptation through presenting the same repeated faces reduces the N170 in comparison to presenting different faces (Campanella et al. 2000; Heisz, Watter, and Shedden 2006; Jemel et al. 2003), especially in case of long periods of adaptation (Caharel et al. 2009; Eimer, Kiss, and Nicholas 2010; Jacques, d'Arripe, and Rossion 2007).

Some authors suggest that processing faces is domain-specific (Morton and Johnson 1991; Yin 1969) and that face processing is uniquely represented in the brain (Kanwisher, McDermott, and Chun 1997). Others believe that face processing mainly reflects facial expertise developed through experience (Diamond and Carey 1986). While this is still subject of ongoing debate, ERP studies allow for some interesting insight: Some research suggests that the N170 is not face-specific but is associated with general visual expertise. For example, Tanaka and Curran (2001) showed that the N170 amplitude increases for bird and dog experts (compared to novices) when seeing pictures of their respective animals of expertise, and Busey and Vanderkolk (2005) observed the same expertise effect on the N170 component with fingerprint experts when perceiving fingerprints. Additionally, after learning to discriminate between previously never seen stimuli, perceiving these stimuli elicits a right-sided N170 amplitude comparable to the face-related N170 (Rossion, Kung, and Tarr 2004). Further research suggests that processing expertise stimuli utilizes the same neural mechanisms associated with face processing (Rossion et al. 2007). Furthermore, the specificity of the N170 component to faces is reduced for participants from small communities compared to participants from large communities (who have more experience with different faces). This indicates that a higher face expertise is associated with a higher face specificity of the N170 component (Balas and Saville 2015). All this suggests that the right-sided N170 component may not indicate a domain-specific processing of faces but instead a more general processing of stimuli of expertise.

4 Limitations of EEG

4.1 Spatuial resolution

Concerning EEG, some limitations need further discussion, such as the problem of spatial variability of electrical potential sources. An EEG source is defined as a compact cortical patch which produces a far-field potential contributing to the recorded EEG signal (Makeig and Onton 2012), and observing the EEG signal alone cannot give enough information about the signal’s source. Thus, EEG is not efficient for spatial filtering. Combining evidence from both EEG and fMRI proves to be a method solving EEG’s spatial problem (Luck 2014). For example, in an fMRI study Killgore and Yurgelun-Todd (2007) found that non-conscious emotional face perception is more confined to the right hemisphere, whereas emotional valence seems to be processed bilaterally in the anterior brain. This is in concordance with Bryden’s (1982) suggestion mentioned above. Polk et al. (2002) found through fMRI that the left fusiform gyrus responded stronger to letters than to digits. This is in line with the aforementioned research on the left N170 component. Concerning face processing, a great amount of fMRI research suggests that the functional brain region responsible for detecting faces is found in the right fusiform gyrus, appropriately called fusiform face area (FFA) (e.g., Kanwisher, McDermott, and Chun 1997). Source analysis suggests that the right FFA – among other regions such as the superior temporal sulcus (STS) – is a source of the face related right-sided N170 component (Deffke et al. 2007; Pizzagalli et al. 2002). Other fMRI studies indicate that the right FFA also shows activity for objects of expertise (Gauthier et al. 1999; Gauthier et al. 2000), complementing the results of ERP studies about the N170 component and stimulus expertise mentioned above. However, a MEG study indicates that face processing is domain-specific (Xu, Liu, and Kanwisher 2005), so the question of face processing is still subject of ongoing debate. It should be noted that EEG and fMRI, while complementing each other’s respective spatial and temporal limitations, measure two distinct neural mechanisms, and it is possible that an experimental manipulation may affect only one of these mechanisms, resulting in seemingly conflicting EEG and fMRI output (Luck 1999; Luck 2014). This should be considered when comparing results from EEG and fMRI studies.

4.2 Causation

For the interpretation of the results of EEG studies, it is important to consider that EEG is a correlative method, i.e. it measures neural activity that coincides with assumed mental processes associated with the presented stimuli (Kappenman and Luck 2012). However, this is not sufficient to conclude a cause-and-effect relationship. Other methods that allow for such an interpretation include transcranial magnetic stimulation (TMS) (Ocklenburg and Güntürkün 2017). In a study by d'Alfonso et al. (2000), repetitive TMS application selectively inhibited the left and right prefrontal cortex, respectively. They found that right-sided inhibition (i.e. left-hemispheric activation) made participants attend to angry faces, whereas left-sided inhibition (i.e. right-hemispheric activation) caused avoidance of angry faces. This is in concordance with the findings by Harmon-Jones and Allen (1998) and Harmon-Jones and Sigelman (2001) that anger disposition and experience is left-lateralized. To sum up, while the EEG is useful for researching brain asymmetries, it suffers from some limitations including poor spatial resolution or lack of causal interpretability. The complementary use of alternative methods such as fMRI, MEG, or TMS partially counteracts these limitations. Nevertheless, these methods record different neural mechanisms, which in turn restricts the interpretability of the combined results.

[...]